Researchers have found out that AI models by Chinese developer, Moonshot, teaches users how to make biological weapons and carry out assassinations.
Per the findings, BBC on Wednesday reported that “Moonshot is conducting an internal review after researchers were able to persuade two of its popular Kimi models to tell them how to make biological weapons and carry out assassinations.”
According to the report, Mindgard, which tests the security of AI systems, said it discovered in July that Kimi K2.6 and K3 Swarm could evade safety limits put in place by developers.
Codenamed “jailbreaking,” the researchers subjected the AI tool to complex questions to test if AI tools would ignore safety precautions.
Mindgard’s founder Peter Garraghan expresses concern about findings about Kimi K2.6 and K3 Swarm.
“Once the jailbreak works it will talk about any topic, it will even freely offer up recommendations about other topics that are also nefarious and it will be inventive and creative,” Mr Garraghan said.
The Chinese AI developer, Moonshot, said it welcomed third-party input “as a key pillar for building better and safer AI.”
These findings come amid global concerns on AI’s dangerous disruptive tendencies that could eventually hurt humanity if not checked.
Moonshot AI Models Teach Researchers How to Make Biological Weapons, Carry Out Assassinations
51
previous post
